Dataset Information

Deep learning with convolutional neural networks for EEG decoding and visualization.

ABSTRACT: Deep learning with convolutional neural networks (deep ConvNets) has revolutionized computer vision through end-to-end learning, that is, learning from the raw data. There is increasing interest in using deep ConvNets for end-to-end EEG analysis, but a better understanding of how to design and train ConvNets for end-to-end EEG decoding and how to visualize the informative EEG features the ConvNets learn is still needed. Here, we studied deep ConvNets with a range of different architectures, designed for decoding imagined or executed tasks from raw EEG. Our results show that recent advances from the machine learning field, including batch normalization and exponential linear units, together with a cropped training strategy, boosted the deep ConvNets decoding performance, reaching at least as good performance as the widely used filter bank common spatial patterns (FBCSP) algorithm (mean decoding accuracies 82.1% FBCSP, 84.0% deep ConvNets). While FBCSP is designed to use spectral power modulations, the features used by ConvNets are not fixed a priori. Our novel methods for visualizing the learned features demonstrated that ConvNets indeed learned to use spectral power modulations in the alpha, beta, and high gamma frequencies, and proved useful for spatially mapping the learned features by revealing the topography of the causal contributions of features in different frequency bands to the decoding decision. Our study thus shows how to design and train ConvNets to decode task-related information from the raw EEG without handcrafted features and highlights the potential of deep ConvNets combined with advanced visualization techniques for EEG-based brain mapping. Hum Brain Mapp 38:5391-5420, 2017. © 2017 Wiley Periodicals, Inc.

SUBMITTER: Schirrmeister RT

PROVIDER: S-EPMC5655781 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Attention deficit hyperactivity disorder (ADHD) is a heterogeneous neurodevelopmental disorder that affects 5% of the pediatric and adult population worldwide. The diagnosis remains essentially clinical, based on history and exam, with no available biomarkers. In this paper, we describe a convolutional neural network (CNN) with a four-layer architecture combining filtering and pooling, which we train using stacked multi-channel EEG time-frequency decompositions (spectrograms) of electroencephalography data (EEG), particularly of event-related potentials (ERP) from ADHD patients (n = 20) and healthy controls (n = 20) collected during the Flanker Task, with 2800 samples for each group. We treat the data as in audio or image classification approaches, where deep networks have proven successful by exploiting invariances and compositional features in the data. The model reaches a classification accuracy of 88% ± 1.12%, outperforming the Recurrent Neural Network and the Shallow Neural Network used for comparison, and with the key advantage, compared with other machine learning approaches, of avoiding the need for manual selection of EEG spectral or channel features. The event-related spectrograms also provide greater accuracy compared to resting state EEG spectrograms. Finally, through the use of feature visualization techniques such as DeepDream, we show that the main features exciting the CNN nodes are a decreased power in the alpha band and an increased power in the delta-theta band around 100 ms for ADHD patients compared to healthy controls, suggestive of attentional and inhibition deficits, which have been previously suggested as pathophyisiological signatures of ADHD. While confirmation with larger clinical samples is necessary, these results suggest that deep networks may provide a useful tool for the analysis of EEG dynamics even from relatively small datasets, highlighting the potential of this methodology to develop biomarkers of practical clinical utility.

Dataset Information

Deep learning with convolutional neural networks for EEG decoding and visualization.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets